ELIZA — the first chatbot

The machine that seemed to understand

Between 1964 and 1966, computer scientist Joseph Weizenbaum at MIT developed ELIZA — a program that could conduct a text conversation with a human by transforming the human's input into questions and reflections. The most famous script, DOCTOR, simulated a Rogerian psychotherapist. ELIZA did not understand language; it used pattern matching and simple transformation rules to create the illusion of understanding.

How ELIZA worked

ELIZA scanned the user's input for keywords from a predefined list. If someone typed "I feel depressed," ELIZA might respond: "Why do you say you feel depressed?" or "How long have you been feeling depressed?" The script contained dozens of such transformation patterns. If no keyword was found, ELIZA used neutral responses such as "Please continue" or "That's very interesting." The illusion of empathy arose not from understanding but from the reflection of the user's own words.

The ELIZA effect

What shocked Weizenbaum most was not how the program worked, but how people reacted to it. Secretaries asked to be left alone with ELIZA, claiming it was private. Weizenbaum's own secretary, who knew it was just a program, asked him to leave the room during her conversation with ELIZA. Psychiatrists seriously suggested ELIZA could replace therapists. This phenomenon — attributing human qualities to a computer program based on superficial signals — became known as the ELIZA effect. It is still one of the most discussed phenomena in human-computer interaction.

Weizenbaum's disillusionment

The reaction to ELIZA disturbed Weizenbaum deeply. In his 1976 book Computer Power and Human Reason, he argued that computers, however capable, should never replace human judgment in matters involving empathy, values, and moral responsibility — such as therapy, law, or education. He became one of the most prominent critics of uncritical AI optimism. His warnings about the ease with which people anthropomorphize machines remain relevant now that millions converse with AI systems daily.

Legacy

ELIZA is the direct ancestor of every chatbot and virtual assistant that followed. It showed that a machine could conduct a conversation that its conversation partners found meaningful, without any real understanding. This laid the foundation for chatbot research, voice assistants, and ultimately large language models. Socially, it warned of the dangers of anthropomorphizing AI — a danger that became more urgent than ever with the rise of ChatGPT and Claude in the 2020s. The questions ELIZA raised about the distinction between simulating and understanding remain unanswered.

Sources

• Weizenbaum, J. (1966). ELIZA — a Computer Program for the Study of Natural Language Communication Between Man and Machine. Communications of the ACM, 9(1), 36–45.
• Weizenbaum, J. (1976). Computer Power and Human Reason: From Judgment to Calculation. W.H. Freeman.
• Turkle, S. (1984). The Second Self: Computers and the Human Spirit. Simon & Schuster.
• Wikipedia — ELIZA
• Original paper (PDF)